The call that cut through the noise

Last week a former Tacton customer called me with three questions. He has led manufacturing P&Ls longer than some product managers have been alive. Fifty minutes in, I realized his questions explained the next decade of CPQ better than any deck I've seen.

I remember running a 700-hour rollout in 2007 and feeling proud we shipped on time. He watched me build a working configurator from a PDF in days and asked, without flinching, "How is that going to change your revenue model?" That was the moment the penny dropped. He was not asking about features. He was asking about economics.

The real shift: economics, not features

Most teams think the shift is about sprinkling a language model on top of an old CPQ stack to shave a few steps. That is not the story. The real story is a collapse in the cost of producing a working configurator. When cost plummets, demand does not stay flat. It expands into the long tail.

Software doesn't just add features. It buys time in bulk.

The translation industry already ran this experiment. Computer-aided translation and then neural networks took translators from 1,000 words a day to 10,000. Everyone predicted a job apocalypse. The market went from roughly 3 billion to more than 70 billion because the backlog of "not worth translating" content became worth it overnight. This is exactly what happens when a cost curve breaks.

CPQ is on that curve now. Projects that took quarters now take days. Model ownership moves from consultants to your own subject matter experts. The unit economics change first. Then the addressable market changes. Then the vendor and services landscape changes. In that order.

Here are the principles that now matter:

  • CPQ is not about automation. It is about correctness. A fast mistake is still a mistake. The constraint solver remains the backbone that guarantees buildable, priceable outcomes.
  • AI does not replace logic. It depends on it. A language model without constraints produces fluent guesses. With explicit, testable rules, it becomes a reasoning assistant.
  • Explainability is trust. If the system cannot explain itself, adoption stalls. Salespeople need to see why a configuration or price is valid.
  • Speed-to-value beats big-project theater. A week of working value is the speedboat. A 600-hour program is the harbor. Buyers want to move now.

How it changes tech, revenue, and your buyer

1) The economic mechanism: "How is that going to change your revenue model?"

Yes, billable hours per project will collapse. But the right lens is not margin protection. It is market creation. When configurators are cheap and fast to stand up, you finally address the long tail: small product lines, regional variants, partner configurators, internal quoting assistants for ops, all the places that never cleared the ROI threshold before.

Expect smaller average deal sizes and far more deals. Expect services to shift from building once to continuous improvement, testing, and enablement. The growth comes from new segments that were always there and priced out. Optimize for the new market, not the old margin.

2) The technical mechanism: "Do you still use a rules engine?"

This sounds like plumbing. It is actually about trust. A pure language model will eventually hallucinate a bill of materials that cannot be built. A pure rules engine will interrogate customers with 47 dropdowns and lose the deal. The winning architecture is hybrid: AI on top of rules.

The language model handles conversation, extraction, and intent. It turns "I need minus 40 degrees, diesel, top inlet" into structured requirements. The constraint solver checks feasibility, resolves trade-offs, and guarantees a correct, explainable configuration. One without the other fails in the real world.

The future of CPQ is hybrid intelligence. Humans set intent, logic ensures correctness, AI accelerates the path between.

Make the rules explicit, modular, and testable. Wrap them with a conversational layer that can explain choices in plain language. If it cannot explain itself, it will not be trusted. Adoption is the only metric that matters.

3) The market mechanism: "Our configurations are simple. Does this fit us?"

For 25 years, the answer was no unless you had painful complexity and patience for a year-long program. That math is gone. When you can stand up a working configurator in a week using a brochure and a subject matter expert, the viable customer profile flips. The "too simple" customer becomes the customer.

What they buy is not an enterprise project. They buy a guided online configurator, a clean quote, engagement tracking, and the ability to maintain it themselves. Think self-checkout, not a staffed counter. Think GPS for complex sales. You say where you need to go, it guides the valid route and avoids dead ends.

Want a quick litmus test? If your buyer can describe the product in a paragraph, you can capture it with a language model and enforce it with rules. If you can enforce it, you can sell it at speed and scale.

What to change this quarter

  • Audit your rules. Are they modular, named, and testable, or tribal and brittle? Rules are not the enemy. Brittle rules are.
  • Run a one-week pilot from a brochure. Prove conversational capture feeding a constraint solver. Demand explainable outcomes and a test suite.
  • Rewrite your buyer profile. Include the segments you previously excluded on cost grounds. Design a speedboat offer for them.
  • Stop waiting for perfect pricing. Ship a good baseline and create learning loops. Perfect pricing is a myth. Learning systems win.

None of this is theory for me. I have lived the heavyweight programs and the speedboats. The teams that win now treat configuration logic as an asset, expose it through a conversational layer, and push value in days, not quarters.

Calm one-sentence truth that sticks: When cost collapses, correctness and explainability decide who keeps the trust and who loses the market.